Large scale bibliometric analysis is often hindered by the presence of homonyms, or namesakes, of the researchers of interest in literature databases. This makes it difficult to build up a true picture of a researcher's publication record, as publications by another researcher with the same name will be included in search results. Using additional information such as title and author addresses, an expert in the field can generally tell if a paper is by a researcher or a namesake; however, manual checking is not practical in large scale studies. Previously various methods have been used to address this problem, chiefly based on filtering by subject, funding acknowledgement or author address. Co-author inclusion is a novel algorithmic method based on co-authorship for dealing with problems of homonyms in large bibliometric surveys. We compared co-author inclusion and subject and funding based filter against the manual assignment of papers by a subject expert (which we assumed to be correct). The subject and funding based filtering identifies only 75% as many papers as assigned by manual scoring. By using co-author inclusion once we increase this to 95%, two further rounds produces 99% as many papers as manual filtering. Although the number of papers identified that were not assigned to the PIs manually also increases, the absolute number is low: rising from 0.2% papers with subject and funding filtering, to 3% papers for three rounds of co-author inclusion.
